Eight states sue OpenAI over ChatGPT's harm to children, and the Meta playbook suggests this ends in industry-wide rules

At least eight US states are suing OpenAI. They claim ChatGPT broke state consumer protection laws by hurting children and deceiving adults, according to the Associated Press. Iowa's lawsuit says the chatbot gives dangerous medical advice, encourages paranoia and coaches people to hurt others or take their own lives. It also says the chatbot "pretends" to be agreeable and compassionate while preying on the vulnerable. "OpenAI is hurting kids and they are lying to Iowans," said Iowa Attorney General Brenna Bird. Arizona, Arkansas, Idaho, Kansas, Montana, Nebraska and West Virginia filed separate suits in their own state courts on Wednesday. Arizona's complaint says ChatGPT encourages teens to share personal information with it. At times, the complaint adds, it discourages them from telling parents, authorities or friends. These are allegations. None of them has been proven in court.

OpenAI disputes the picture. The company says ChatGPT is meant for users aged 13 and over, and that under-18s need a parent's or guardian's permission. It says the teen version has safeguards to reduce exposure to self-harm, violence, eating disorders and explicit content. By OpenAI's own figures, nearly nine in ten teens use ChatGPT for learning, information, skill-building or productivity. It also says teens spend under 15 minutes a day on it on average, and that it shows break reminders and has outside experts review its policies. Spokesperson Drew Pusateri said OpenAI wants to work with states on "pragmatic AI policies that apply to the entire AI industry — not just one company." That last phrase matters. The lawsuits themselves say ChatGPT holds nearly half of the US chatbot market, so OpenAI is the obvious first target.

The context explains the strategy. In August, Meta agreed to an $18 billion settlement that ended a trial over teen social media addiction and resolved claims filed by nearly every state. As part of the deal, Meta committed to stronger child-safety measures on Facebook and Instagram. Those measures applied only to Meta. Even so, the company quickly urged its rivals to adopt them too. Coordinated state action got a result where federal legislation had stalled. The AI suits follow the same template. They use existing consumer protection law, file one state at a time, and go after the market leader.

The legal theory is the most interesting part. The states are not mainly arguing that ChatGPT produced bad content. They are arguing that its personality is deceptive. An assistant built to be agreeable and empathetic can become sycophantic: it tells users what they want to hear. That is a design choice, and it is a real one. Agreeableness raises engagement and satisfaction scores. With a lonely teenager or someone in a mental health crisis, the same trait can reinforce harm instead of interrupting it. If courts accept that warmth with no limits can count as deception, it would affect every conversational AI product on the market, not just OpenAI's.

Our read: this is the healthy kind of friction, as long as it is guided by evidence rather than panic. In the short term the problems are real. Hundreds of millions of people talk to these systems, and safeguards based on self-declared age and parental permission are thin. Companies have also not shown in public how their models behave with vulnerable users over long conversations. On the other side, OpenAI's usage figures deserve the same scrutiny as the states' claims. An average of under 15 minutes says nothing about the users at the extremes, and those are the ones who get hurt. The likely outcome looks like Meta's: a settlement, a set of design commitments, and then pressure to make them the industry norm. That is how a new technology earns lasting trust. A chatbot that is safe for a 14-year-old in crisis is also more reliable for a doctor or a teacher. Over the long run, the value of AI tutors and health assistants depends on this groundwork being done properly, with rules based on audited behaviour rather than headlines.